Visualization of risk and uncertainty to communicate avalanche hazard: An Empirical Study
2Fabrikant, S.I.; 3Streit, A.; 3Purves, R.S.
Avalanche accidents are, in comparison to the number of individuals active in the mountains, relatively rare events. Nevertheless, in the last 20 years ~3500 people were involved in ~1900 avalanche accidents in Switzerland. Some 90% of these were undertaking recreational activities in uncontrolled terrain (Techel and Zweifel, 2013). In such terrain, parties must make their own decisions about the practicality of a route on the basis of a wide range of factors, including the terrain they intend to visit, current avalanche hazard and snow stability, prevailing and expected meteorological conditions, experience and skills of the party and personal goals. Avalanche education seeks to provide tools to help structure the decision making process, for example through the so-called 3x3 method which considers three factors (conditions, terrain and human factors) at three levels (planning, in the terrain, and on individual slopes) (Munter 2007; Harvey and Nigg, 2009). One tool provided to ease the decision making process, particularly at the planning stage, is the graphic reduction method, Here, two-dimensional colored grid, where avalanche danger levels as communicated through the daily avalanche bulleting are read off the x-axis, and steepness of slope is indicated on the y-axis (i.e., < 30% | 30–35 | 36–40 | >40 degrees), uses a traffic light metaphor to make initial recommendations and fuzzy borders to communicate uncertainty. Using the method requires transferring information contained therein to a 1:25000 topographic map when planning a route. Avalanche risk assessment thus requires both accurate reading and comprehension of a graph and accurate visual extraction of slope information from a topographic map for the planned route and their meaningful integration. We report on an empirical study where we propose a map design solution for trip planning showing avalanche risk zones computed by the graphic reduction method and integrated into a topographic map. This solution is based the visual variables location, crispness, and color value, shown by MacEachren et al., (2012) to be well understood by map users when assessing uncertainty data at ordinal scales. In a within-subject study, 19 participants (9 females, 10 males) were asked to assess avalanche risk with the current graphical reduction method (where the transfer to the map was made by the participants) and with our newly proposed integrated map design solution. Our results suggest that participants’ response accuracy is significantly higher (F(1,17)) = 5.390, p = 0.033, < .05) with our newly proposed intrinsic risk visualization method (M= 94.3%, SD=8.4), compared to the graphic reduction method (M= 88,4%, SD=10). This performance difference is also supported by participants’ significantly higher average response confidence ratings (F(1,18)) = 5.179, p = 0.035, < .05) for the new intrinsic risk map (M= 1.93), compared to the graphical reduction method, (M= 2.19), assessed on a 5-point Likert scale (1= very confident to 5 = not confident at all). Also, noteworthy is the good performance of participants at a complex task. There has been a long-standing debate in the visualization literature whether or not to show data uncertainty together with the data, and if so, how (MacEachren, 2005). Empirical assessment of intrinsic uncertainty visualization solutions have been scarce (MacEachren et al. 2012), especially in the context of a real world decision making scenarios involving risk. Our results suggest that the direct integration of uncertainty information into a map increases participants’ response accuracy and confidence using the graphic reduction method. However, the generally high ability of participants to transfer from the graphic reduction method to a map also suggests that the present method may already be of considerable value.